Construction Morphology and the interaction of syntax and word formation*
Geert Booij (University of Leiden)
1. Introduction
The relation between word formation and syntax has been a persistent topic of research in
the publications of my esteemed colleague professor Soledad Varela. Therefore, I hope it
is a proper tribute to her work on morphology to deal with certain aspects of this kind of
interaction in my contribution to this volume on the occasion of her retirement.
Lexical units can be constructed in two ways, by means of word formation
(morphological operations) and by syntax, as also pointed out in Rainer & Varela (1992)
in their discussion of the demarcation of compounds and phrases in Spanish. One obvious
function of word formation mechanisms such as compounding and derivation is
extending the set of lexical units of a language. However, the set of lexical units of a
language is larger than the set of words. Phrasal units of various size form lexical units
as well, varying from small phrases to complete sentences, and the construction of such
units contribute to the growth and the size of the lexicon as well. It is therefore important
not to identify the notions ‘word’ and ‘lexical unit’. Simplex words are lexical units by
definition, and complex words are lexical units when they are conventionalized (‘existing
complex words’). Many complex words never make it to the status of lexical unit since
they are not conventionalized, and are used only once. Hence, expanding the lexicon is
only one function of word formation. Its other function is to provide relatively compact
linguistic expressions for denoting all kinds of concepts.
In this paper, I will discuss various types of interaction between word formation
and syntax. In particular, I will show that paradigmatic relationships between phrasal and
morphological schemas for the construction of lexical units play an important role. In
section 2, I summarize the evidence for phrasal lexical units, a notion that we need for a
proper analysis of the interaction of syntax and word formation in the construction of
lexical units. Section 3 presents some basic notions of Construction Grammar that are
presupposed in my analysis. The following two sections discuss a number of specific
cases of morphology-syntax interaction. I will argue that these phenomena provide
evidence for an architecture of the grammar in which there is no strict division between
grammar and lexicon, and in which networks of paradigmatic relations between linguistic
expressions play a crucial role in a proper account of the form and meaning of lexical
units.
2. The pervasiveness of prefabs
In the linguistic literature lexical phrasal units are referred to as phraseologisms
(Fleischer 1992; 1997) or prefabs (Erman & Warren 2000). A well known class of such
lexical phrases are combinations of an adjective and a noun. Compare the following A+N
phrases in English and their glosses in Dutch:
(1)
English
Dutch
a.
b.
strong tea / *mighty tea
sterke thee ‘strong tea’
weak tea / *feeble tea
slappe thee ‘lit. slack tea’
a steady girlfriend
een vaste vriendin ‘lit. a fixed girlfriend’
a confirmed bachelor
een verstokte vrijgezel ‘lit. a hardened
bachelor’
an eligible bachelor
een begerenswaardige vrijgezel ‘lit. a
desirable bachelor’
The English examples in (1a) are given by Erman and Warren (2000), who point out that
there is a fixed choice of adjective in the examples (1a): neither can strong be replaced
with mighty, nor weak with feeble. The Dutch glosses exhibit the same type of
conventionality. In the case of strong tea, Dutch requires the direct translation of strong,
the adjective sterk, but in the case of weak tea, the direct translation zwak cannot be used,
and instead the adjective slap ‘loose’ is the proper choice. The same observation can be
made for the English phrases and their Dutch glosses in (1b). Similarly, in a detailed
study of English verbs, Dąbrowska concluded that “speakers have very specific
knowledge about the collocations and semantic preferences of individual verbs”
(Dąbrowska 2009).
The examples (1) illustrate that knowledge of the lexical units of a language
comprises much more than knowledge of its vocabulary. Knowledge of conventional
word combinations is essential for a full mastery of a language. In using a non-native
language, we might speak grammatically correct, but un-idiomatically. Acquiring the
competence of idiomaticity is therefore an important goal of second language education
(Boers et al. 2006). Special dictionaries of word combinations, such as Benson et al.
(1997) for English, are very useful for achieving this goal.
The term ‘prefab’ used by Erman and Warren (2000) for this type of lexical unit
is more adequate than the term ‘idiom’, because we tend to associate the latter term with
the idea of non-compositionality: an idiom is usually assumed to possess a (partially)
unpredictable meaning. The term ‘fixed expression’ is a good alternative (Sprenger 2003).
The term ‘prefab’ rightly indicates that we have to do here with ready-made multi-word
chunks that reflect the conventionality of language use. For example, there is no semantic
opacity involved in the meaning of the English phrase strong tea (compare this phrase to
the A+N phrases red army and yellow fever that do have idiomatic meanings). The
specific knowledge involved in the case of strong tea is only that of the selection of the
adjective strong rather than mighty. These kinds of collocations can therefore be qualified
as ‘idioms of encoding’ (Gagné & Spalding 2006) since they do not pose a problem for
decoding their meaning, but specify how a particular meaning should be encoded.
Knowledge of prefabs should be an important component of models of language
production and perception (Kuiper et al. 2007; Sprenger 2003; Sprenger et al. 2006).
Erman and Warren (2000) estimate that about 55 % of the utterances in written English
consist of prefabs. Prefab knowledge is very important for language production because it
is crucial in the task of lexical choice as part of the encoding process. Psycholinguistic
evidence confirms the psychological reality and relevance of prefabs since phrasal
sequences exhibit frequency effects, just like words, even when these phrases are
compositional (Tremblay & Baayen 2010).
Knowledge of prefabs has received renewed attention in the linguistic literature,
because its implications for our view of the architecture of the grammar, and of language
change (Booij 2010; Bybee & Beckner 2010; Bybee & Torres Cacoullos 2009; Ernestus
& Baayen 2011; Jackendoff 1997; 2002a; 2002b; Wray 2002; 2008).
Below, I will argue that the theory of Construction Grammar (CxG) and its
application to the construction of lexical units, Construction Morphology (CM), provide
an insightful account of the role and position of lexical units, including prefabs, in the
grammars of natural languages.
3. Construction Grammar
In CxG both complex words and phrasal lexical units are constructs, that is, pairings of
forms and meanings. Sets of linguistic constructs with corresponding, recurrent properties
can be characterized by means of abstract schemas. Consider the English A+N phrases
strong tea and red army mentioned above. They have the regular structure [A N]NP, and
hence are instantiations of the English Noun Phrase. A central idea of CxG is that both
schemas and their instantiations are stored in the grammar. An abstract schema dominates
its individual instantiations, which inherit by the properties defined in the schema. Thus,
the predictable properties of individual AN phrases count as redundant information. This
makes it possible to store completely regular but conventionalized phrases in the lexicon
without obliterating their regular character.3 This view is also defended by Jackendoff in
his Parallel Architecture (PA) approach to grammar (Jackendoff 2002a; 2010; 2011).
There are two main reasons for storing specific phrasal patterns. One reason is
that they are constructions in the sense of (Goldberg 1995: 4), that is, they may have
semantic properties that cannot be predicted from the meaning contribution of the
constituent words. The other reason is that instantiations of phrasal patterns may be
entrenched; therefore, we need a phrasal schema that generalizes over the common
properties of entrenched phrases such as AN phrases (Goldberg 2006: 5), Zeschel (2009).
Jackendoff (2010: 224) argues that we should allow for schemas that do not have a
specific constructional meaning, such as the schemas for English AN phrases, particle
verbs, and NN compounds. One may, however, maintain the position that what is stored
are always constructions, since regular phrasal patterns can always be assigned a very
general constructional meaning. In the case of English AN phrases, for instance, the
constructional meaning is that of ‘modifier – modified’. That is, the construction imposes
a modifier interpretation on the adjective, and takes the head noun to denote the modified
concept.
In the remainder of this article, I will therefore assume the correctness of the
assumption that both abstract schemas for lexical (syntactic and morphological) units,
and their conventionalized instantiations are stored in the grammar. This can be modeled
as a hierarchical lexicon, with constructions at different levels of expressions (Booij
2010). For Dutch compounds, for instance, we need sub-schemas because constituents of
compounds may feature a specific ‘bound’ meaning when embedded in compounds. An
example of this phenomenon is the set of words with intensifying meaning that occur as
the left constituents of Dutch XA compounds, where X = N, A or V:
(2)
Intensifying lexemes in Dutch X A compounds
X = noun
examples
ber-e‘bear’
bere-sterk ‘very strong’, bere-gezellig ‘very cosy’
bloed ‘blood’
bloed-serieus ‘very serious’, bloed-link ‘very risky’
X = adjective
dol ‘mad’
dol-blij ‘very happy’, dol-gelukkig ‘very happy’
stom ‘stupid’
stom-toevallig ‘completely coincidental’, stom-verbaasd
‘very surprised’
X = verb
kots ‘vomit’
kots-misselijk ‘very sick’, kots-beu ‘very tired of’
loei ‘sizzle’
loei-hard ‘very hard’, loei-goed ‘very good’
In these words, the first constituent no longer carries its lexical meaning as specified in
the left column, but functions as a word with an abstract intensifier meaning. In a few
cases there might still be a semantic explanation for the intensifier interpretation of the
modifying noun and the head: bere-sterk, for instance, can be interpreted as ‘as strong as
a bear’. However, there is no obvious way in which the word gezellig ‘cosy’ can be
related to a typical property of bears. This semantic reanalysis of such words as
intensifiers is made overt by the fact that they can be attached productively with this
specific intensifier meaning to form new adjectives. A noteworthy point concerning the
intensifiers bere-, rete-, and reuze- is that they consist of a consonant-final stem followed
by a linking element -e [ə]. They appear without schwa when used as independent words.
This linking element is a necessary part of these nouns when used as intensifiers.
We can formally express the affixoid nature of these compound-initial lexemes by
specifying them in constructional idioms of the following form:
(3)
[[bere]Nk [x]Ai]Aj ↔ [veryk SEMi]j
[[dol]Ak[x]Ai]A j ↔ [veryk SEMi]j
[[loei]Vk[x]Ai]A j ↔ [veryk SEMi]j
The symbol ↔ stands for the correspondence relation between form and meaning. On the
left hand side we have a form specification, and on the right hand side a specification of
the corresponding meaning. The semantic contribution of specific formal sub-constituents
is indicated by means of co-indexation (Jackendoff 2002a). Constructional idioms are
morphological or syntactic schemas in which one or more positions are lexically fixed,
whereas other positions are open slots, represented by variables (Jackendoff 2002a).
Being embedded in constructional schemas makes these words similar to affixes. The
difference is that affixes do not carry a lexical category label, and hence cannot be related
to independent lexemes in the lexicon. The schemas (3) specify that the words on the left
do not have their regular lexical meaning but evoke, in these morphological constructions,
the meaning of intensification (‘to a high degree’) or excellence.
4. Lexical phrases and names
In probably all European languages we find phrases with coordination that are
semantically transparent, but have a fixed order, such as English salt and pepper, father
and son, ladies and gentlemen (Malkiel 1959). These conventionalized expressions have
to be listed because of the fixed order in which the two words appear. On the other hand,
they instantiate regular and productive phrasal patterns, and we should express this in our
description. Therefore, the syntactic coordination schema [N and N]NP, which is a subcase of English coordination, dominates its instantiations, phrases such as salt and pepper.
The only non-redundant information concerning these ! and ! phrases is that they exist
(that is, are conventionalized), and the order in which the two words appear.
Thus, there is no principled difference between complex words and phrases with respect
to the division of labour between storage and computation. Storage of complex words is
necessary, one reason being that we have to specify which possible complex words are
conventional, that is actual words.
A correct prediction of the position that both words and phrases are stored as
lexical units is that we will find competition and blocking effects between synonymous
phrases and words. For instance, the existence of the Dutch AN compound zuur-kool
‘sour-cabbage, Sauerkraut’ impedes using the AN phrase zure kool ‘sour cabbage’ for
this type of cabbage, and inversely, the existence of the AN phrase rode kool ‘red
cabbage’ impedes the coinage of the AN compound rood-kool. Related languages may
differ in the choice between synonymous but structurally different options. For instance,
where German has a very productive pattern of AN compounds, Dutch speakers tend to
prefer AN phrases (as in German Rotwein ‘red wine’ versus Dutch rode wijn ‘red wine’)
(Booij 2002). Romance languages also use a lot of NA phrases as names for concepts.
Phrasal lexical units may also arise through a special form of syntax. For instance,
names for functions, organizations, etc. may exhibit ‘headline syntax’, the type of syntax
that is allowed in the headlines of newspapers, with omission of grammatical words, as
illustrated by the following example:
(4)
director cultural affairs
In this construct, the head director is on the left. Hence it cannot be a compound, since
English compounds are right-headed. On the other hand, it is not a regular phrase since in
NPs the complement must be governed by a preposition, which is lacking here. Hence, it
is a case of special syntax, allowed in headlines of newspapers, but also in names for
functions (Booij 2009b). This implies the assumption of special phrasal schemas for
extending the English fund of lexical expressions for functions and institutions:
(5)
[.. Ni NPj]NPk ↔ [SEMi with relation R to SEMj]k SEMi = function, institution, ...
5. Form-meaning asymmetries
As we saw above, phrases may form building blocks of complex words. But what
happens if a word formation process does not take phrases as bases for affixation?
Consider the following examples from Italian:
(6)
chirarra elettrica ‘electric guitar’
chitarrista elettrico ‘electrical guitarist’
flauto barocco ‘baroque flute’
flautista barocco ‘baroque flute player’
tennis da tavola ‘tennis table’
tennista da tavolo ‘table tennis player’
The word sequences on the left and the right are NPs of the form A + N or N Prep N. The
phrase chitarrista elettrico denotes someone who plays the electric guitar. However, the
phrase chitarra elettrica ‘lit. guitar electric, electric guitar’ is not the formal base for the
attachment of the suffix -ista that is used in Italian to create personal names. This is clear
from the word order (the suffix is not attached at the right edge, *chitarra elettricista),
and from the fact that the adjective elettrico agrees in number and gender with the
masculine noun chitarrista, and not with the feminine noun chitarra. This asymmetry
between form and meaning (‘bracketing paradox’) is the same as that in the famous
English example transformational grammarian discussed in Spencer (1988). In the case
of English, one might consider the suffix -ian to be formally attached to the phrase
transformational grammar, an option that is available due to the English word order and
the poor inflection of English. Such an analysis is obviously impossible in the case of
Italian, which only allows for a phrasal interpretation:
(7)
[[chitarr-ista]N [elettrico]A]NP
This suggests that the adjective elettrico has semantic scope over a part of the complex
phrasal head guitarrista, and that the internal morphological structure of the modified
head word is accessible. In example (7), the adjective elettrico modifies the nominal base
of the head noun. In the following examples AN phrases used as modifiers are turned into
AA sequences, in order to fit the canonical phrase structure of Italian (examples provided
to me by Daniele Vergilitto):
(8)
energia solare ‘solar energy’
impiantoN energeticoA solareA ‘solar
power plant’
chirurgiaN esteticaA ‘aesthetic surgery’
interventoN chirurgicoA esteticoA
‘aesthetic surgery invention’
Let us look at these facts from the perspective of how to encode meaning. When
an Italian speaker knows the lexical phrase guitarra elettrica, and he wants to construct
the lexical unit for denoting a person playing on an electric guitar, the suffix -ista that is
used for creating personal names cannot be attached to the phrase guitarra elettrica
because this is not a word. The noun *gitarra elettric-ista is ill formed. The alternative is
a phrase with a form-meaning mismatch in which the (stem forms of the) relevant
lexemes (guitarra and elettrico) that form a lexical collocation, are inserted in the
available slots in a schema that complies with the restriction that -ista does not accept
phrasal bases. This is the schema [[N-ista]N A]NP. The semantic property to be specified
is that the Adjective has semantic scope over the N only.
Productive patterns with this kind of asymmetry such as the Italian one can be
accounted for by specifying a paradigmatic relationship between two phrasal schemas
(Booij 2010: 140), where ≈ denotes a paradigmatic relationship:
(9)
< [Nj Ak]NPi ↔ SEMi > ≈ < [ [Nj-ista]N Ak]NPl ↔ [PERSON with relation R to
[SEMi]]l >
6. Conclusions
In this article it was shown that syntax and morphology may interact in a number of ways
in order to expand the set of names for concepts in a language. The naming function is
not the exclusive domain of word formation. Phrases also function as names, and in
addition they form bases from which names can be derived in special ways such as the
formation of acronyms and stump compounds.
The use of phrases as units of name formation is increased by languages allowing
for asymmetries between formal structure and semantic structure. Such asymmetries
appeared not to form a problem in encoding or decoding, thanks to the availability of
paradigmatic relationships between linguistic expressions.
These observations about the various relationships between phrases and complex
words that play a role in constructing and expanding the lexicon of a language find a
natural and insightful account in the theory of Construction Grammar, which makes use
of phrasal and morphological schemas. These schemas can be unified, thus giving rise to
complex words with phrases as building blocks. By specifying both abstract schemas and
their instantiations, we can do justice to both the creative and the conventional aspect of
linguistic knowledge. Moreover, schemas can be related paradigmatically, which
provides the flexibility in coining lexical units that is needed.
*The research for this paper has been supported by a fellowship of the Alexander von
Humboldt Foundation, which is gratefully acknowledged here.
References
Benson, Morton, Evelyn Benson & Robert Ilson. 1997. The BBI disctionary of English
word combinations Amsterdam / Philadelphia: Benjamins.
Boers, Frank, June Eyckmans, Jenny Kappel, Hélène Stengers & Murielle Demecheleer.
2006. Formulaic sequences and perceived oral proficiency: Putting a lexical
approach to the test. Language Teaching Research 10.245-61.
Booij, Geert. 2010. Construction morphology Oxford: Oxford University Press.
Bybee, Joan & Clay Beckner. 2010. Usage-based theory. The Oxford handbook of
linguistic analysis, ed. by B. Heine & H. Narrog, 827-56. Oxford: Oxford
University Press.
Bybee, Joan & Rena Torres Cacoullos. 2009. The role of prefabs in grammaticization.
How the particular and the general interact in language change. Formulaic
language. Vol 1. Distribution and historical change, ed. by R.L. Corrigan, E.
Moravcsik, H. Ouali & K.M. Wheatly, 195-217. Amsterdam / Philadelphia:
Benjamins.
Dąbrowska, Ewa. 2009. Words as constructions. New directions in Cognitive Linguistics,
ed. by V. Evans & S. Pourcel, 201-23. Amsterdam / Philadelphia: Benjamins.
Erman, Britt & Beatrice Warren. 2000. The idiom principle and the open choice principle.
Text 20.29-62.
Ernestus, Mirjam & R. Harald Baayen. 2011. Corpora and exemplars in phonology.
Handbook of phonology, ed. by A. Yu & J. Goldsmith. Oxford: Blackwell.
Fleischer, Wolfgang. 1992. Konvergenz und Divergenz von Wortbildung und
Phraseologisierung. Phraseologie und Wordbildung. Aspekte der
Lexikonerweiterung, ed. by J. Korhonen, 53-65. Tübingen: Niemeyer.
Fleischer, Wolfgang. 1997. Phraseologie der deutschen Gegenwartssprache. 2e,
erweiterte Auflage Tübingen: Niemeyer.
Gagné, Christine L. & Thomas L. Spalding. 2006. Conceptual combination: implication
for the mental lexicon. The representation and processing of compound words, ed.
by G. Libben & G. Jarema, 145-68. Oxford: Oxford University Press.
Goldberg, Adele. 1995. Constructions. A Construction Grammar approach to argument
structure Chicago: Chicago University Press.
Goldberg, Adele. 2006. Constructions at work. The nature of generalization in language
Oxford: Oxford University Press.
Jackendoff, Ray. 1997. The architecture of the language faculty Cambridge Mass.: MIT
Press.
Jackendoff, Ray. 2002a. Foundations of language Oxford: Oxford University Press.
Jackendoff, Ray. 2002b. What's in the lexicon? Storage and computation in the language
faculty, ed. by S. Nooteboom, F. Weerman & F. Wijnen, 3-40. Dordrecht: Kluwer.
Jackendoff, Ray. 2010. Meaning and the lexicon. The parallel architecture 1975-2010
Oxford: Oxford University Press.
Jackendoff, Ray. 2011. Formal CxG: Constructions in the Parallel Architecture. The
Oxford handbook of Construction Grammar, ed. by T. Hoffmann & G. Trousdale.
Oxford: Oxford University Press.
Kuiper, Koenraad, Marie-Elaine Van Egmond, Gerard Kempen & Simone Sprenger.
2007. Slipping on superlemmas. Multi-word lexical items in speech production.
The Mental Lexicon 2.313-57.
Rainer, Franz & Soledad Varela. 1992. Compounding in Spanish. Rivista di Linguistica
4.117-42.
Sprenger, Simone A. 2003. Fixed expressions and the production of idioms Nijmegen:
Max-Planck-Institut für Psycholinguisitk.
Sprenger, Simone, Willem J. M. Levelt & Gerard Kempen. 2006. Lexical access during
the production of idiomatic phrases. Journal of Memory and Language 54.161-84.
Tremblay, A. & R. Harald Baayen. 2010. Holistic processing of regular four-word
sequences: A behavioral and erp study of the effects of structure, frequency, and
probability on immediate free recall. Perspectives on formulaic language.
Acquisition and communication, ed. by D. Wood, 151-73. London: The
Continuum International Publishing Group.
Wray, Alison. 2002. Formulaic language and the lexicon Cambridge: Cambridge
University Press.
Wray, Alison. 2008. Formulaic language: pushing the boundaries Oxford: Oxford
University Press.
Zeschel, Arne. 2009. What's (in) a construction? Complete inheritance vs. full-entry
models. New directions in Cognitive Linguistics, ed. by V. Evans & S. Pourcel,
185-200. Amsterdam / Philadelphia: Benjamins.